SlideShare a Scribd company logo
1 of 63
ReviewReview
 Types of dataTypes of data
 p-valuep-value
 Steps for hypothesis testSteps for hypothesis test
– How do we set up a null hypothesis?How do we set up a null hypothesis?
 Choosing the right testChoosing the right test
– Continuous outcome variable/dichotomousContinuous outcome variable/dichotomous
explanatory variable: Two sample t-testexplanatory variable: Two sample t-test
Steps for hypothesis testingSteps for hypothesis testing
1)1) State null hypothesisState null hypothesis
2)2) State type of data for explanatory and outcomeState type of data for explanatory and outcome
variablevariable
3)3) Determine appropriate statistical testDetermine appropriate statistical test
4)4) State summary statisticsState summary statistics
5)5) Calculate p-value (stat package)Calculate p-value (stat package)
6)6) Decide whether to reject or not reject the nullDecide whether to reject or not reject the null
hypothesishypothesis
• NEVER accept nullNEVER accept null
1)1) Write conclusionWrite conclusion
ExampleExample
 In previous class, two groups wereIn previous class, two groups were
compared on a continuous outcomecompared on a continuous outcome
 What if we have more than two groups?What if we have more than two groups?
 Ex. A recent study compared the intensityEx. A recent study compared the intensity
of structures on MRI in normal controls,of structures on MRI in normal controls,
benign MS patients and secondarybenign MS patients and secondary
progressive MS patientsprogressive MS patients
 Question: Is there any difference amongQuestion: Is there any difference among
these groups?these groups?
Two approachesTwo approaches
 Compare each group to each other groupCompare each group to each other group
using a t-testusing a t-test
– Problem withProblem with multiple comparisonsmultiple comparisons
 CompleteComplete global comparisonglobal comparison to see ifto see if
there is any differencethere is any difference
– Analysis of variance (ANOVA)Analysis of variance (ANOVA)
– Good first step even if eventually completeGood first step even if eventually complete
pairwise comparisonspairwise comparisons
Types of analysis-independentTypes of analysis-independent
samplessamples
OutcomeOutcome ExplanatoryExplanatory AnalysisAnalysis
ContinuousContinuous DichotomousDichotomous t-test, Wilcoxont-test, Wilcoxon
testtest
ContinuousContinuous CategoricalCategorical ANOVA, linearANOVA, linear
regressionregression
ContinuousContinuous ContinuousContinuous Correlation, linearCorrelation, linear
regressionregression
DichotomousDichotomous DichotomousDichotomous Chi-square test,Chi-square test,
logistic regressionlogistic regression
DichotomousDichotomous ContinuousContinuous Logistic regressionLogistic regression
Time to eventTime to event DichotomousDichotomous Log-rank testLog-rank test
Global test-ANOVAGlobal test-ANOVA
 As a first step, we can compare across allAs a first step, we can compare across all
groups at oncegroups at once
 The null hypothesis for ANOVA is that theThe null hypothesis for ANOVA is that the
means in all of the groups are equalmeans in all of the groups are equal
 ANOVA compares the within groupANOVA compares the within group
variance and the between group variancevariance and the between group variance
– If the patients within a group are very alikeIf the patients within a group are very alike
and the groups are very different, the groupsand the groups are very different, the groups
are likely differentare likely different
Hypothesis testHypothesis test
1)1) HH00: mean: meannormalnormal=mean=meanBMSBMS=mean=meanSPMSSPMS
2)2) Outcome variable: continuousOutcome variable: continuous
Explanatory variable: categoricalExplanatory variable: categorical
3)3) Test: ANOVATest: ANOVA
4)4) meanmeannormalnormal=0.41; mean=0.41; meanBMSBMS= 0.34; mean= 0.34; meanSPMSSPMS=0.30=0.30
5)5) Results: p=0.011Results: p=0.011
6)6) Reject null hypothesisReject null hypothesis
7)7) Conclusion: At least one of the groups isConclusion: At least one of the groups is
significantly different than the otherssignificantly different than the others
Technical asideTechnical aside
 Our F-statistic is the ratio of the between groupOur F-statistic is the ratio of the between group
variance and the within group variancevariance and the within group variance
 This ratio of variances has a known distribution (F-This ratio of variances has a known distribution (F-
distribution)distribution)
 If our calculated F-statistic is high, the between groupIf our calculated F-statistic is high, the between group
variance is higher than the within group variance,variance is higher than the within group variance,
meaning the differences between the groups are notmeaning the differences between the groups are not
likely due to chancelikely due to chance
 Therefore, the probability of the observed result orTherefore, the probability of the observed result or
something more extreme will be low (low p-value)something more extreme will be low (low p-value)
( )
( ) ( )( ) ( ) ( )( )1111
1
1
22
11
1
2
2
2
−++−−++−
−−
==
∑=
kkk
k
i
ii
within
between
nnsnsn
kxxn
s
s
F

This is the
distribution under the
null
This small shaded
region is the part of
the distribution that is
equal to or more
extreme than the
observed value.
The p-value!!!
Now whatNow what
 The question often becomes which groupsThe question often becomes which groups
are differentare different
 Possible comparisonsPossible comparisons
– All pairsAll pairs
– All groups to a specific controlAll groups to a specific control
– Pre-specified comparisonsPre-specified comparisons
 If we do many tests, we should account forIf we do many tests, we should account for
multiple comparisonsmultiple comparisons
Type I errorType I error
 Type I error is when you reject the nullType I error is when you reject the null
hypothesis even though it is truehypothesis even though it is true
((αα=P(reject H=P(reject H00|H|H00 is true))is true))
 We accept making this error 5% of theWe accept making this error 5% of the
timetime
 If we run a large experiment with 100 testsIf we run a large experiment with 100 tests
and the null hypothesis was true in eachand the null hypothesis was true in each
case, how many times would we expect tocase, how many times would we expect to
reject the null?reject the null?
Multiple comparisonsMultiple comparisons
 For this problem, three comparisonsFor this problem, three comparisons
– NC vs. BMS; NC vs. SPMS; BMS vs. SPMSNC vs. BMS; NC vs. SPMS; BMS vs. SPMS
 If we complete each test at the 0.05 level, whatIf we complete each test at the 0.05 level, what
is the chance that we make a type I error?is the chance that we make a type I error?
– P(reject at least 1 | HP(reject at least 1 | H00 is true)is true) == αα
– P(reject at least 1 | HP(reject at least 1 | H00 is true)is true) = 1-= 1- P(fail to reject allP(fail to reject all
three| Hthree| H00 is true)is true) = 1-0.95= 1-0.9533
= 0.143= 0.143
 Inflated type I error rateInflated type I error rate
 Can correct p-value for each test to maintainCan correct p-value for each test to maintain
experiment type I errorexperiment type I error
Bonferroni correctionBonferroni correction
 TheThe Bonferroni correctionBonferroni correction multiples all p-multiples all p-
values by the number of comparisons completedvalues by the number of comparisons completed
– In our experiment, there were 3 comparisons, so weIn our experiment, there were 3 comparisons, so we
multiply by 3multiply by 3
– Any p-value that remains less than 0.05 is significantAny p-value that remains less than 0.05 is significant
 The Bonferroni correction is conservative (it isThe Bonferroni correction is conservative (it is
more difficult to obtain a significant result than itmore difficult to obtain a significant result than it
should be), but it is an extremely easy way toshould be), but it is an extremely easy way to
account for multiple comparisons.account for multiple comparisons.
– Can be very harsh correction with many testsCan be very harsh correction with many tests
Other correctionsOther corrections
 All pairwise comparisonsAll pairwise comparisons
– Tukey’s testTukey’s test
 All groups to a controlAll groups to a control
– Dunnett’s testDunnett’s test
 MANY othersMANY others
 False discovery rateFalse discovery rate
ExampleExample
 For our three-group comparison, we compareFor our three-group comparison, we compare
each and get the following results from Tukey’seach and get the following results from Tukey’s
testtest
GroupsGroups Mean diffMean diff p-valuep-value SignificantSignificant
NC vs. BMSNC vs. BMS 0.0750.075 0.100.10
NC vs. SPMSNC vs. SPMS 0.1140.114 0.0120.012 **
BMS vs.BMS vs.
SPMSSPMS
0.0390.039 0.600.60
Questions to ask yourselfQuestions to ask yourself
 What is the null hypothesis?What is the null hypothesis?
 We would like to test the null hypothesis atWe would like to test the null hypothesis at
the 0.05 levelthe 0.05 level
 If well defined prior to the experiment, theIf well defined prior to the experiment, the
correction for multiple comparison ifcorrection for multiple comparison if
necessary will be clearnecessary will be clear
 Hypothesis generating vs.Hypothesis generating vs.
hypothesis testinghypothesis testing
ConclusionsConclusions
 If you are doing a multiple group comparison,If you are doing a multiple group comparison,
always specify before the experiment whichalways specify before the experiment which
comparisons are of interest if possiblecomparisons are of interest if possible
 If the null hypothesis is that all the groups areIf the null hypothesis is that all the groups are
the same, test global null using ANOVAthe same, test global null using ANOVA
 Complete appropriate additional comparisonsComplete appropriate additional comparisons
with corrections if necessarywith corrections if necessary
 No single right answer for every situationNo single right answer for every situation
Types of analysis-independentTypes of analysis-independent
samplessamples
OutcomeOutcome ExplanatoryExplanatory AnalysisAnalysis
ContinuousContinuous DichotomousDichotomous t-test, Wilcoxont-test, Wilcoxon
testtest
ContinuousContinuous CategoricalCategorical ANOVA, linearANOVA, linear
regressionregression
ContinuousContinuous ContinuousContinuous Correlation, linearCorrelation, linear
regressionregression
DichotomousDichotomous DichotomousDichotomous Chi-square test,Chi-square test,
logistic regressionlogistic regression
DichotomousDichotomous ContinuousContinuous Logistic regressionLogistic regression
Time to eventTime to event DichotomousDichotomous Log-rank testLog-rank test
CorrelationCorrelation
 Is there a linearIs there a linear
relationship betweenrelationship between
IL-10 expression andIL-10 expression and
IL-6 expression?IL-6 expression?
 The best graphicalThe best graphical
display for this data isdisplay for this data is
a scatter plota scatter plot
CorrelationCorrelation
 DefinitionDefinition: the degree to which two continuous: the degree to which two continuous
variables are linearly relatedvariables are linearly related
– Positive correlation- As one variable goes up, thePositive correlation- As one variable goes up, the
other goes up (positive slope)other goes up (positive slope)
– Negative correlation- As one variable goes up, theNegative correlation- As one variable goes up, the
other goes down (negative slope)other goes down (negative slope)
 Correlation (Correlation (ρρ) ranges from -1 (perfect negative) ranges from -1 (perfect negative
correlation) to 1 (perfect positive correlation)correlation) to 1 (perfect positive correlation)
 A correlation of 0 means that there is no linearA correlation of 0 means that there is no linear
relationship between the two variablesrelationship between the two variables
Positive correlation
0
2
4
6
8
10
12
0 2 4 6 8 10 12
Negative correlation
0
2
4
6
8
10
12
0 2 4 6 8 10 12
No correlation
0
1
2
3
4
5
6
7
8
9
10
0 2 4 6 8 10 12
No correlation (quadratic)
0
2
4
6
8
10
12
14
16
18
0 2 4 6 8 10
Hypothesis testHypothesis test
1)1) HH00: correlation between IL-10 expression and: correlation between IL-10 expression and
IL-6 expression=0IL-6 expression=0
2)2) Outcome variable: IL-6 expression- continuousOutcome variable: IL-6 expression- continuous
Explanatory variable: IL-10 expression-Explanatory variable: IL-10 expression-
continuouscontinuous
3)3) Test: correlationTest: correlation
4)4) Summary statistic: correlation=0.51Summary statistic: correlation=0.51
5)5) Results: p=0.011Results: p=0.011
6)6) Reject null hypothesisReject null hypothesis
7)7) Conclusion: A statistically significantConclusion: A statistically significant
correlation was observed between the twocorrelation was observed between the two
variablesvariables
Technical aside-correlationTechnical aside-correlation
 The formal definition of the correlation is given by:The formal definition of the correlation is given by:
 Note that this is dimensionless quantityNote that this is dimensionless quantity
 This equation shows that if the covariance between theThis equation shows that if the covariance between the
two variables is the same as the variance in the twotwo variables is the same as the variance in the two
variables, we have perfect correlation because all of thevariables, we have perfect correlation because all of the
variability in x and y is explained by how the twovariability in x and y is explained by how the two
variables change togethervariables change together
)()(
),(
),(
yVarxVar
yxCov
yxCorr =
How can we estimate theHow can we estimate the
correlation?correlation?
 The most common estimator of the correlation is theThe most common estimator of the correlation is the
Pearson’s correlation coefficientPearson’s correlation coefficient, given by:, given by:
 This is a estimate that requires both x and y are normallyThis is a estimate that requires both x and y are normally
distributed. Since we use the mean in the calculation, thedistributed. Since we use the mean in the calculation, the
estimate is sensitive to outliers.estimate is sensitive to outliers.
( )( )
( ) ( ) 





−





−
−−
=
∑∑
∑
==
=
n
i
i
n
i
i
n
i
ii
yyxx
yyxx
r
1
2
1
2
1
Distribution of the test statisticDistribution of the test statistic
 The standard error of the sample correlationThe standard error of the sample correlation
coefficient is given bycoefficient is given by
 The resulting distribution of the test statistic is a t-The resulting distribution of the test statistic is a t-
distribution with n-2 degrees of freedom where ndistribution with n-2 degrees of freedom where n
is the number of patients (not the number ofis the number of patients (not the number of
measurements)measurements)
2
1
)(ˆ
2
−
−
=
n
r
res
22 1
2
2
1
0
r
n
r
n
r
r
t
−
−
=
−
−
−
=
Regression-Everything in one placeRegression-Everything in one place
 All analyses we have done to this pointAll analyses we have done to this point
can be completed using regression!!!can be completed using regression!!!
Quick math reviewQuick math review
 As you remember, theAs you remember, the
equation of a line isequation of a line is
y=mx+by=mx+b
 FFor every one unitor every one unit
increase in x, there isincrease in x, there is
an m unit increase inan m unit increase in
yy
 bb is the value of yis the value of y
when x is equal towhen x is equal to
zerozero
Line
y = 1.5x + 4
0
2
4
6
8
10
12
14
16
18
20
0 2 4 6 8 10 12
PicturePicture
 Does there seem toDoes there seem to
be a linearbe a linear
relationship in therelationship in the
data?data?
 Is the data perfectlyIs the data perfectly
linear?linear?
 Could we fit a line toCould we fit a line to
this data?this data?
0
5
10
15
20
25
0 2 4 6 8 10 12
How do we find the best line?How do we find the best line?
 Linear regressionLinear regression
tries to find the besttries to find the best
line (curve) to fit theline (curve) to fit the
data Let’s look atdata Let’s look at
three candidate linesthree candidate lines
 Which do you think isWhich do you think is
the best?the best?
 What is a way toWhat is a way to
determine the bestdetermine the best
line to use?line to use?
What is linear regression?What is linear regression?
 The method of findingThe method of finding
the best line (curve) isthe best line (curve) is
least squares, whichleast squares, which
minimizes theminimizes the
distance from the linedistance from the line
for each of pointsfor each of points
 The equation of theThe equation of the
line is y=1.5x + 4line is y=1.5x + 4
y = 1.5x + 4
0
5
10
15
20
25
0 2 4 6 8 10 12
ExampleExample
 For our investigation of theFor our investigation of the
relationship between IL-10relationship between IL-10
and IL-6, we can set up aand IL-6, we can set up a
regression equationregression equation
 ββ00 is the expression of IL-6is the expression of IL-6
when IL-10=0 (intercept)when IL-10=0 (intercept)
 ββ11 is the change in IL-6 foris the change in IL-6 for
every 1 unit increase in IL-10every 1 unit increase in IL-10
(slope)(slope)
 εεii is the residual from the lineis the residual from the line
iii ILIL εββ ++= 10*6 10
 The final regression equation isThe final regression equation is
 The coefficients meanThe coefficients mean
– the estimate of the mean expression of IL-6 for athe estimate of the mean expression of IL-6 for a
patient with IL-10 expression=0 (patient with IL-10 expression=0 (ββ00))
– an increase of one unit in IL-10 expression leads toan increase of one unit in IL-10 expression leads to
an estimated increase of 0.63 in the meanan estimated increase of 0.63 in the mean
expression of IL-6 (expression of IL-6 (ββ11))
10*63.04.266ˆ ILIL +=
Tough questionTough question
 In our correlation hypothesis test, we wanted toIn our correlation hypothesis test, we wanted to
know if there was an association between theknow if there was an association between the
two measurestwo measures
 If there was no relationship between IL-10 andIf there was no relationship between IL-10 and
IL-6 in our system, what would happen to ourIL-6 in our system, what would happen to our
regression equation?regression equation?
– No effect means that the change in IL-6 is not relatedNo effect means that the change in IL-6 is not related
to the change in IL-10to the change in IL-10
– ββ11=0=0
 IsIs ββ11 significantly different than zero?significantly different than zero?
Hypothesis testHypothesis test
1)1) HH00: no relationship between IL-6 expression: no relationship between IL-6 expression
and IL-10 expression,and IL-10 expression, ββ11 =0=0
2)2) Outcome variable: IL-6- continuousOutcome variable: IL-6- continuous
Explanatory variable: IL-10- continuousExplanatory variable: IL-10- continuous
3)3) Test: linear regressionTest: linear regression
4)4) Summary statistic:Summary statistic: ββ11 = 0.63= 0.63
5)5) Results: p=0.011Results: p=0.011
6)6) Reject null hypothesisReject null hypothesis
7)7) Conclusion: A significant correlation wasConclusion: A significant correlation was
observed between the two variablesobserved between the two variables
Wait a second!!Wait a second!!
 Let’s check somethingLet’s check something
– p-value from correlation analysis = 0.011p-value from correlation analysis = 0.011
– p-value from regression analysis = 0.011p-value from regression analysis = 0.011
– They are the same!!They are the same!!
 Regression leads to same conclusion asRegression leads to same conclusion as
correlation analysiscorrelation analysis
 Other similarities as well from modelsOther similarities as well from models
Technical aside-Estimates ofTechnical aside-Estimates of
regression coefficientsregression coefficients
 Once we have solved the least squaresOnce we have solved the least squares
equation, we obtain estimates for theequation, we obtain estimates for the ββ’s, which’s, which
we refer to aswe refer to as
 To test if this estimate is significantly differentTo test if this estimate is significantly different
than 0, we use the following equation:than 0, we use the following equation:
10
ˆ,ˆ ββ
( )( )
( )
xy
xx
yyxx
n
i
i
n
i
ii
10
1
2
1
1
ˆˆ
ˆ
ββ
β
−=
−
−−
=
∑
∑
=
=
( )1
11
ˆˆ
ˆ
β
ββ
es
t
−
=
Assumptions of linear regressionAssumptions of linear regression
 LinearityLinearity
– Linear relationship between outcome and predictorsLinear relationship between outcome and predictors
– E(Y|X=x)=E(Y|X=x)=ββ00 ++ ββ11xx11 ++ ββ22xx22
22
is still a linear regressionis still a linear regression
equation because each of theequation because each of the ββ’s is to the first power’s is to the first power
 Normality of the residualsNormality of the residuals
– The residuals,The residuals, εεii, are normally distributed, N(0,, are normally distributed, N(0, σσ22
))
 Homoscedasticity of the residualsHomoscedasticity of the residuals
– The residuals,The residuals, εεii, have the same variance, have the same variance
 IndependenceIndependence
– All of the data points are independentAll of the data points are independent
– Correlated data points can be taken into accountCorrelated data points can be taken into account
using multivariate and longitudinal data methodsusing multivariate and longitudinal data methods
Linear regression with dichotomousLinear regression with dichotomous
predictorpredictor
 Linear regression can also be used forLinear regression can also be used for
dichotomous predictors, like sexdichotomous predictors, like sex
 Last class we compared relapsing MS patientsLast class we compared relapsing MS patients
to progressive MS patientsto progressive MS patients
 To do this, we use an indicator variable, whichTo do this, we use an indicator variable, which
equals 1 for relapsing and 0 for progressive. Theequals 1 for relapsing and 0 for progressive. The
resulting regression equation for expression isresulting regression equation for expression is
iii Rex εββ ++= *10
Interpretation of modelInterpretation of model
 The meaning of the coefficients in this case areThe meaning of the coefficients in this case are
– ββ00 is the estimate of the mean expression whenis the estimate of the mean expression when
R=0, in the progressive groupR=0, in the progressive group
– ββ00 + β+ β11 is the estimate of the mean expression whenis the estimate of the mean expression when
R=1, in the relapsing groupR=1, in the relapsing group
– ββ11 is the estimate of the mean increase in expressionis the estimate of the mean increase in expression
between the two groupsbetween the two groups
 The difference between the two groups isThe difference between the two groups is ββ11
 If there was no difference between the groups,If there was no difference between the groups,
what wouldwhat would ββ11 equal?equal?
Mean in wildtype=β0
Mean in
Progressive
group=β0
Difference between
groups=β1
Hypothesis testHypothesis test
1)1) Null hypothesis: meanNull hypothesis: meanprogressiveprogressive=mean=meanrelapsingrelapsing ((ββ11=0)=0)
2)2) Explanatory: group membership- dichotomousExplanatory: group membership- dichotomous
Outcome: cytokine production-continuousOutcome: cytokine production-continuous
3)3) Test: Linear regressionTest: Linear regression
4)4) ββ11=6.87=6.87
5)5) p-value=0.199p-value=0.199
6)6) Fail to reject null hypothesisFail to reject null hypothesis
7)7) Conclusion: The difference between theConclusion: The difference between the
groups is not statistically significantgroups is not statistically significant
T-testT-test
 As hopefully you remember, you couldAs hopefully you remember, you could
have tested this same null hypothesishave tested this same null hypothesis
using a two sample t-testusing a two sample t-test
 Very similar result to previous classVery similar result to previous class
 If we would have assumed equal varianceIf we would have assumed equal variance
for our t-test, we would have gotten to thefor our t-test, we would have gotten to the
same result!!!same result!!!
 ANOVA results can also be tested usingANOVA results can also be tested using
regression using more than one indicatorregression using more than one indicator
Multiple regressionMultiple regression
 A large advantage of regression is the ability toA large advantage of regression is the ability to
include multiple predictors of an outcome in oneinclude multiple predictors of an outcome in one
analysisanalysis
 A multiple regression equation looks just like aA multiple regression equation looks just like a
simple regression equation.simple regression equation.
exxxY nn +++++= ββββ ...22110
ExampleExample
 Brain parenchymal fraction (BPF) is aBrain parenchymal fraction (BPF) is a
measure of disease severity in MSmeasure of disease severity in MS
 We would like to know if gender has anWe would like to know if gender has an
effect on BPF in MS patientseffect on BPF in MS patients
 We also know that BPF declines with ageWe also know that BPF declines with age
in MS patientsin MS patients
 Is there an effect of sex on BPF if weIs there an effect of sex on BPF if we
control for age?control for age?
.75.8.85.9.95
BPF
0 .2 .4 .6 .8 1
Sex
Blue=males; Red=females
Blue=males; Red=females
.75.8.85.9.95
BPF
20 30 40 50 60
Age
Is age a potential confounder?Is age a potential confounder?
 We know that age has an effect on BPFWe know that age has an effect on BPF
from previous researchfrom previous research
 We also know that male patients have aWe also know that male patients have a
different disease course than femaledifferent disease course than female
patients so the age at time of samplingpatients so the age at time of sampling
may also be related to sexmay also be related to sex
BPFSex
Age
ModelModel
 The multiple linear regression modelThe multiple linear regression model
includes a term for both age and sexincludes a term for both age and sex
 What are the values genderWhat are the values genderii takes on?takes on?
– gendergenderii=0 if the patient is female=0 if the patient is female
– gendergenderii=1 if the patient is male=1 if the patient is male
iiii agegenderBPF εβββ +++= ** 210
ExpressionExpression
 Females:Females:
– BPFBPFii == ββ00++ ββ22*age*ageii++εεii
 Males:Males:
– BPFBPFii = (= (ββ00++ ββ11)+)+ ββ22*age*ageii++εεii
 What is different about the equations?What is different about the equations?
– InterceptIntercept
 What is the same?What is the same?
– SlopeSlope
 This model allows an effect of gender on theThis model allows an effect of gender on the
intercept, but not on the change with ageintercept, but not on the change with age
 The meaning of each coefficientThe meaning of each coefficient
– ββ00:: the average BPF when age is 0 and the patient isthe average BPF when age is 0 and the patient is
femalefemale
– ββ11:: the average difference in BPF between males andthe average difference in BPF between males and
female, HOLDING AGE CONSTANTfemale, HOLDING AGE CONSTANT
– ββ22:: the average increase in BPF for a one unitthe average increase in BPF for a one unit
increase in age, HOLDING GENDER CONSTANTincrease in age, HOLDING GENDER CONSTANT
 Note that the interpretation of the coefficientNote that the interpretation of the coefficient
requires mention of the other variables in therequires mention of the other variables in the
modelmodel
Interpretation of coefficientsInterpretation of coefficients
Estimated coefficientsEstimated coefficients
 Here is the estimated regression equationHere is the estimated regression equation
 The average difference between males andThe average difference between males and
females is 0.017 holding age constantfemales is 0.017 holding age constant
 For every one unit increase in age, the meanFor every one unit increase in age, the mean
BPF decreases 0.0026 units holding sex constantBPF decreases 0.0026 units holding sex constant
 Are either of these effects statistically significant?Are either of these effects statistically significant?
– What is the null hypothesis?What is the null hypothesis?
iii agesexFBP *0026.0*017.0942.0ˆ −+=
Hypothesis testHypothesis test
1)1) HH00: No effect of sex, controlling for age: No effect of sex, controlling for age ββ11 =0=0
2)2) Continuous outcome, continuous predictorContinuous outcome, continuous predictor
3)3) Linear regression controlling for sexLinear regression controlling for sex
4)4) Summary statistic:Summary statistic: ββ11 =0.017=0.017
5)5) p-value=0.37p-value=0.37
6)6) Since the p-value is more than 0.05, we fail toSince the p-value is more than 0.05, we fail to
reject the null hypothesisreject the null hypothesis
7)7) We conclude that there is no significantWe conclude that there is no significant
association between sex and BPF controllingassociation between sex and BPF controlling
for agefor age
Hypothesis testHypothesis test
1)1) HH00: No effect of age, controlling for sex: No effect of age, controlling for sex ββ22 =0=0
2)2) Continuous outcome, continuous predictorContinuous outcome, continuous predictor
3)3) Linear regression controlling for sexLinear regression controlling for sex
4)4) Summary statistic:Summary statistic: ββ22 =-0.0026=-0.0026
5)5) p-value=0.00p-value=0.00 44
6)6) Since the p-value is less than 0.05, we rejectSince the p-value is less than 0.05, we reject
the null hypothesisthe null hypothesis
7)7) We conclude that there is a significantWe conclude that there is a significant
association between age and BPF controllingassociation between age and BPF controlling
for sexfor sex
Estimated effect
of age
p-value for age
Estimated effect
of sex
p-value for sex
.75.8.85.9.95
BPF
20 30 40 50 60
Age
ConclusionsConclusions
 Although there was a marginallyAlthough there was a marginally
significant association of sex and BPF,significant association of sex and BPF,
this association was not significant afterthis association was not significant after
controlling for agecontrolling for age
 The significant association between ageThe significant association between age
and BPF remained statistically significantand BPF remained statistically significant
after controlling for sexafter controlling for sex
What we learned (hopefully)What we learned (hopefully)
 ANOVAANOVA
 CorrelationCorrelation
 Basics of regressionBasics of regression

More Related Content

What's hot

Hypothesis testing an introduction
Hypothesis testing an introductionHypothesis testing an introduction
Hypothesis testing an introductionGeetika Gulyani
 
Basis of statistical inference
Basis of statistical inferenceBasis of statistical inference
Basis of statistical inferencezahidacademy
 
6 estimation hypothesis testing t test
6 estimation hypothesis testing t test6 estimation hypothesis testing t test
6 estimation hypothesis testing t testPenny Jiang
 
Hypothesis testing
Hypothesis testingHypothesis testing
Hypothesis testingSumit Sharma
 
Testing of hypotheses
Testing of hypothesesTesting of hypotheses
Testing of hypothesesRajThakuri
 
Chapter 8 – Hypothesis Testing
Chapter 8 – Hypothesis TestingChapter 8 – Hypothesis Testing
Chapter 8 – Hypothesis Testingguest3720ca
 
hypothesis testing-tests of proportions and variances in six sigma
hypothesis testing-tests of proportions and variances in six sigmahypothesis testing-tests of proportions and variances in six sigma
hypothesis testing-tests of proportions and variances in six sigmavdheerajk
 
Chap#9 hypothesis testing (3)
Chap#9 hypothesis testing (3)Chap#9 hypothesis testing (3)
Chap#9 hypothesis testing (3)shafi khan
 
Presentation on Hypothesis Test by Ashik Amin Prem
Presentation on Hypothesis Test by Ashik Amin PremPresentation on Hypothesis Test by Ashik Amin Prem
Presentation on Hypothesis Test by Ashik Amin PremAshikAminPrem
 

What's hot (20)

Hypothesis testing
Hypothesis testingHypothesis testing
Hypothesis testing
 
Hypothesis testing an introduction
Hypothesis testing an introductionHypothesis testing an introduction
Hypothesis testing an introduction
 
Testing of hypothesis
Testing of hypothesisTesting of hypothesis
Testing of hypothesis
 
Basis of statistical inference
Basis of statistical inferenceBasis of statistical inference
Basis of statistical inference
 
6 estimation hypothesis testing t test
6 estimation hypothesis testing t test6 estimation hypothesis testing t test
6 estimation hypothesis testing t test
 
Hypothesis testing
Hypothesis testingHypothesis testing
Hypothesis testing
 
More Statistics
More StatisticsMore Statistics
More Statistics
 
Hypothesis
HypothesisHypothesis
Hypothesis
 
Hypothesis testing
Hypothesis testingHypothesis testing
Hypothesis testing
 
Testing of hypotheses
Testing of hypothesesTesting of hypotheses
Testing of hypotheses
 
Hypothesis Testing
Hypothesis TestingHypothesis Testing
Hypothesis Testing
 
Hypothesis Testing
Hypothesis TestingHypothesis Testing
Hypothesis Testing
 
Hypothesis testing
Hypothesis testingHypothesis testing
Hypothesis testing
 
Chapter 8 – Hypothesis Testing
Chapter 8 – Hypothesis TestingChapter 8 – Hypothesis Testing
Chapter 8 – Hypothesis Testing
 
Hypothesis test
Hypothesis testHypothesis test
Hypothesis test
 
Hypothesis Testing
Hypothesis TestingHypothesis Testing
Hypothesis Testing
 
hypothesis testing-tests of proportions and variances in six sigma
hypothesis testing-tests of proportions and variances in six sigmahypothesis testing-tests of proportions and variances in six sigma
hypothesis testing-tests of proportions and variances in six sigma
 
Chap#9 hypothesis testing (3)
Chap#9 hypothesis testing (3)Chap#9 hypothesis testing (3)
Chap#9 hypothesis testing (3)
 
Presentation on Hypothesis Test by Ashik Amin Prem
Presentation on Hypothesis Test by Ashik Amin PremPresentation on Hypothesis Test by Ashik Amin Prem
Presentation on Hypothesis Test by Ashik Amin Prem
 
Basics of Hypothesis Testing
Basics of Hypothesis Testing  Basics of Hypothesis Testing
Basics of Hypothesis Testing
 

Viewers also liked

Ketut swarjana epidemiology measurement
Ketut swarjana epidemiology measurementKetut swarjana epidemiology measurement
Ketut swarjana epidemiology measurementktswarjana
 
Biostats epidemiological studies
Biostats epidemiological studiesBiostats epidemiological studies
Biostats epidemiological studiesJagdish Dukre
 
XNN001 Introductory epidemiological concepts - Study design
XNN001 Introductory epidemiological concepts - Study designXNN001 Introductory epidemiological concepts - Study design
XNN001 Introductory epidemiological concepts - Study designramseyr
 
Lesson 6 Nonparametric Test 2009 Ta
Lesson 6 Nonparametric Test 2009 TaLesson 6 Nonparametric Test 2009 Ta
Lesson 6 Nonparametric Test 2009 TaSumit Prajapati
 
parametric hypothesis testing using MATLAB
parametric hypothesis testing using MATLABparametric hypothesis testing using MATLAB
parametric hypothesis testing using MATLABKajal Saraswat
 
6. Randomised controlled trial
6. Randomised controlled trial6. Randomised controlled trial
6. Randomised controlled trialRazif Shahril
 
Clinical Research Methodology
Clinical  Research  MethodologyClinical  Research  Methodology
Clinical Research Methodologydrmomusa
 
Test of hypothesis
Test of hypothesisTest of hypothesis
Test of hypothesisJaspreet1192
 
Brm (one tailed and two tailed hypothesis)
Brm (one tailed and two tailed hypothesis)Brm (one tailed and two tailed hypothesis)
Brm (one tailed and two tailed hypothesis)Upama Dwivedi
 
Randomised Controlled Trials
Randomised Controlled TrialsRandomised Controlled Trials
Randomised Controlled Trialsfondas vakalis
 
Randomised controlled trials
Randomised controlled trialsRandomised controlled trials
Randomised controlled trialsHesham Gaber
 
Is a parametric or nonparametric method appropriate with relationship-oriente...
Is a parametric or nonparametric method appropriate with relationship-oriente...Is a parametric or nonparametric method appropriate with relationship-oriente...
Is a parametric or nonparametric method appropriate with relationship-oriente...Ken Plummer
 
Clinical research ppt,
Clinical research   ppt,Clinical research   ppt,
Clinical research ppt,Malay Singh
 
Parametric vs Nonparametric Tests: When to use which
Parametric vs Nonparametric Tests: When to use whichParametric vs Nonparametric Tests: When to use which
Parametric vs Nonparametric Tests: When to use whichGönenç Dalgıç
 
Hypothesis testing examples on z test
Hypothesis testing examples on z testHypothesis testing examples on z test
Hypothesis testing examples on z testJags Jagdish
 
Hypothesis Testing-Z-Test
Hypothesis Testing-Z-TestHypothesis Testing-Z-Test
Hypothesis Testing-Z-TestRoger Binschus
 
Test of hypothesis
Test of hypothesisTest of hypothesis
Test of hypothesisvikramlawand
 

Viewers also liked (19)

Research methodology simplified
Research methodology simplifiedResearch methodology simplified
Research methodology simplified
 
Ketut swarjana epidemiology measurement
Ketut swarjana epidemiology measurementKetut swarjana epidemiology measurement
Ketut swarjana epidemiology measurement
 
Biostats epidemiological studies
Biostats epidemiological studiesBiostats epidemiological studies
Biostats epidemiological studies
 
XNN001 Introductory epidemiological concepts - Study design
XNN001 Introductory epidemiological concepts - Study designXNN001 Introductory epidemiological concepts - Study design
XNN001 Introductory epidemiological concepts - Study design
 
Lesson 6 Nonparametric Test 2009 Ta
Lesson 6 Nonparametric Test 2009 TaLesson 6 Nonparametric Test 2009 Ta
Lesson 6 Nonparametric Test 2009 Ta
 
parametric hypothesis testing using MATLAB
parametric hypothesis testing using MATLABparametric hypothesis testing using MATLAB
parametric hypothesis testing using MATLAB
 
6. Randomised controlled trial
6. Randomised controlled trial6. Randomised controlled trial
6. Randomised controlled trial
 
Clinical Research Methodology
Clinical  Research  MethodologyClinical  Research  Methodology
Clinical Research Methodology
 
Test of hypothesis
Test of hypothesisTest of hypothesis
Test of hypothesis
 
Brm (one tailed and two tailed hypothesis)
Brm (one tailed and two tailed hypothesis)Brm (one tailed and two tailed hypothesis)
Brm (one tailed and two tailed hypothesis)
 
Randomised Controlled Trials
Randomised Controlled TrialsRandomised Controlled Trials
Randomised Controlled Trials
 
Randomised controlled trials
Randomised controlled trialsRandomised controlled trials
Randomised controlled trials
 
Is a parametric or nonparametric method appropriate with relationship-oriente...
Is a parametric or nonparametric method appropriate with relationship-oriente...Is a parametric or nonparametric method appropriate with relationship-oriente...
Is a parametric or nonparametric method appropriate with relationship-oriente...
 
Clinical research ppt,
Clinical research   ppt,Clinical research   ppt,
Clinical research ppt,
 
Parametric vs Nonparametric Tests: When to use which
Parametric vs Nonparametric Tests: When to use whichParametric vs Nonparametric Tests: When to use which
Parametric vs Nonparametric Tests: When to use which
 
Hypothesis testing examples on z test
Hypothesis testing examples on z testHypothesis testing examples on z test
Hypothesis testing examples on z test
 
Z test
Z testZ test
Z test
 
Hypothesis Testing-Z-Test
Hypothesis Testing-Z-TestHypothesis Testing-Z-Test
Hypothesis Testing-Z-Test
 
Test of hypothesis
Test of hypothesisTest of hypothesis
Test of hypothesis
 

Similar to Math3010 week 5

Hypothesis Testing in Six Sigma
Hypothesis Testing in Six SigmaHypothesis Testing in Six Sigma
Hypothesis Testing in Six SigmaBody of Knowledge
 
Review Z Test Ci 1
Review Z Test Ci 1Review Z Test Ci 1
Review Z Test Ci 1shoffma5
 
Statistical inference: Statistical Power, ANOVA, and Post Hoc tests
Statistical inference: Statistical Power, ANOVA, and Post Hoc testsStatistical inference: Statistical Power, ANOVA, and Post Hoc tests
Statistical inference: Statistical Power, ANOVA, and Post Hoc testsEugene Yan Ziyou
 
Steps of hypothesis testingSelect the appropriate testSo far.docx
Steps of hypothesis testingSelect the appropriate testSo far.docxSteps of hypothesis testingSelect the appropriate testSo far.docx
Steps of hypothesis testingSelect the appropriate testSo far.docxdessiechisomjj4
 
PAGE O&M Statistics – Inferential Statistics Hypothesis Test.docx
PAGE  O&M Statistics – Inferential Statistics Hypothesis Test.docxPAGE  O&M Statistics – Inferential Statistics Hypothesis Test.docx
PAGE O&M Statistics – Inferential Statistics Hypothesis Test.docxgerardkortney
 
Data science classica_hypos
Data science classica_hyposData science classica_hypos
Data science classica_hyposNeeraj Sinha
 
Hypothesis Tests in R Programming
Hypothesis Tests in R ProgrammingHypothesis Tests in R Programming
Hypothesis Tests in R ProgrammingAtacan Garip
 
Assessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docxAssessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docxgalerussel59292
 
Assessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docxAssessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docxfestockton
 
Hypothesis testing1
Hypothesis testing1Hypothesis testing1
Hypothesis testing1HanaaBayomy
 
10. sampling and hypotehsis
10. sampling and hypotehsis10. sampling and hypotehsis
10. sampling and hypotehsisKaran Kukreja
 
Statistical Significance Tests.pptx
Statistical Significance Tests.pptxStatistical Significance Tests.pptx
Statistical Significance Tests.pptxAldofChrist
 
Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric)
Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric) Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric)
Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric) Dexlab Analytics
 
Unit 4 Tests of Significance
Unit 4 Tests of SignificanceUnit 4 Tests of Significance
Unit 4 Tests of SignificanceRai University
 
What So Funny About Proportion Testv3
What So Funny About Proportion Testv3What So Funny About Proportion Testv3
What So Funny About Proportion Testv3ChrisConnors
 

Similar to Math3010 week 5 (20)

Hypothesis Testing in Six Sigma
Hypothesis Testing in Six SigmaHypothesis Testing in Six Sigma
Hypothesis Testing in Six Sigma
 
Review Z Test Ci 1
Review Z Test Ci 1Review Z Test Ci 1
Review Z Test Ci 1
 
Statistical inference: Statistical Power, ANOVA, and Post Hoc tests
Statistical inference: Statistical Power, ANOVA, and Post Hoc testsStatistical inference: Statistical Power, ANOVA, and Post Hoc tests
Statistical inference: Statistical Power, ANOVA, and Post Hoc tests
 
Steps of hypothesis testingSelect the appropriate testSo far.docx
Steps of hypothesis testingSelect the appropriate testSo far.docxSteps of hypothesis testingSelect the appropriate testSo far.docx
Steps of hypothesis testingSelect the appropriate testSo far.docx
 
L hypo testing
L hypo testingL hypo testing
L hypo testing
 
PAGE O&M Statistics – Inferential Statistics Hypothesis Test.docx
PAGE  O&M Statistics – Inferential Statistics Hypothesis Test.docxPAGE  O&M Statistics – Inferential Statistics Hypothesis Test.docx
PAGE O&M Statistics – Inferential Statistics Hypothesis Test.docx
 
Spss session 1 and 2
Spss session 1 and 2Spss session 1 and 2
Spss session 1 and 2
 
Data science classica_hypos
Data science classica_hyposData science classica_hypos
Data science classica_hypos
 
Tests of significance
Tests of significanceTests of significance
Tests of significance
 
Hypothesis Tests in R Programming
Hypothesis Tests in R ProgrammingHypothesis Tests in R Programming
Hypothesis Tests in R Programming
 
Assessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docxAssessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docx
 
Assessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docxAssessment 4 ContextRecall that null hypothesis tests are of.docx
Assessment 4 ContextRecall that null hypothesis tests are of.docx
 
Hypothesis testing
Hypothesis testingHypothesis testing
Hypothesis testing
 
Hypothesis testing1
Hypothesis testing1Hypothesis testing1
Hypothesis testing1
 
10. sampling and hypotehsis
10. sampling and hypotehsis10. sampling and hypotehsis
10. sampling and hypotehsis
 
Stat topics
Stat topicsStat topics
Stat topics
 
Statistical Significance Tests.pptx
Statistical Significance Tests.pptxStatistical Significance Tests.pptx
Statistical Significance Tests.pptx
 
Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric)
Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric) Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric)
Basic of Statistical Inference Part-V: Types of Hypothesis Test (Parametric)
 
Unit 4 Tests of Significance
Unit 4 Tests of SignificanceUnit 4 Tests of Significance
Unit 4 Tests of Significance
 
What So Funny About Proportion Testv3
What So Funny About Proportion Testv3What So Funny About Proportion Testv3
What So Funny About Proportion Testv3
 

More from stanbridge

Micro Lab 3 Lecture
Micro Lab 3 LectureMicro Lab 3 Lecture
Micro Lab 3 Lecturestanbridge
 
Creating a poster v2
Creating a poster v2Creating a poster v2
Creating a poster v2stanbridge
 
Creating a poster
Creating a posterCreating a poster
Creating a posterstanbridge
 
OT 5018 Thesis Dissemination
OT 5018 Thesis DisseminationOT 5018 Thesis Dissemination
OT 5018 Thesis Disseminationstanbridge
 
Ot5101 005 week 5
Ot5101 005 week 5Ot5101 005 week 5
Ot5101 005 week 5stanbridge
 
Ot5101 005 week4
Ot5101 005 week4Ot5101 005 week4
Ot5101 005 week4stanbridge
 
Compliance, motivation, and health behaviors
Compliance, motivation, and health behaviors Compliance, motivation, and health behaviors
Compliance, motivation, and health behaviors stanbridge
 
Ch 5 developmental stages of the learner
Ch 5   developmental stages of the learnerCh 5   developmental stages of the learner
Ch 5 developmental stages of the learnerstanbridge
 
OT 5101 week2 theory policy
OT 5101 week2 theory policyOT 5101 week2 theory policy
OT 5101 week2 theory policystanbridge
 
OT 5101 week3 planning needs assessment
OT 5101 week3 planning needs assessmentOT 5101 week3 planning needs assessment
OT 5101 week3 planning needs assessmentstanbridge
 
NUR 304 Chapter005
NUR 304 Chapter005NUR 304 Chapter005
NUR 304 Chapter005stanbridge
 
NUR 3043 Chapter007
NUR 3043 Chapter007NUR 3043 Chapter007
NUR 3043 Chapter007stanbridge
 
NUR 3043 Chapter006
NUR 3043 Chapter006NUR 3043 Chapter006
NUR 3043 Chapter006stanbridge
 
NUR 3043 Chapter004
NUR 3043 Chapter004NUR 3043 Chapter004
NUR 3043 Chapter004stanbridge
 
3043 Chapter009
3043 Chapter0093043 Chapter009
3043 Chapter009stanbridge
 
3043 Chapter008
 3043 Chapter008 3043 Chapter008
3043 Chapter008stanbridge
 
Melnyk ppt chapter_21
Melnyk ppt chapter_21Melnyk ppt chapter_21
Melnyk ppt chapter_21stanbridge
 
Melnyk ppt chapter_22
Melnyk ppt chapter_22Melnyk ppt chapter_22
Melnyk ppt chapter_22stanbridge
 

More from stanbridge (20)

Micro Lab 3 Lecture
Micro Lab 3 LectureMicro Lab 3 Lecture
Micro Lab 3 Lecture
 
Creating a poster v2
Creating a poster v2Creating a poster v2
Creating a poster v2
 
Creating a poster
Creating a posterCreating a poster
Creating a poster
 
Sample poster
Sample posterSample poster
Sample poster
 
OT 5018 Thesis Dissemination
OT 5018 Thesis DisseminationOT 5018 Thesis Dissemination
OT 5018 Thesis Dissemination
 
Ot5101 005 week 5
Ot5101 005 week 5Ot5101 005 week 5
Ot5101 005 week 5
 
Ot5101 005 week4
Ot5101 005 week4Ot5101 005 week4
Ot5101 005 week4
 
Compliance, motivation, and health behaviors
Compliance, motivation, and health behaviors Compliance, motivation, and health behaviors
Compliance, motivation, and health behaviors
 
Ch 5 developmental stages of the learner
Ch 5   developmental stages of the learnerCh 5   developmental stages of the learner
Ch 5 developmental stages of the learner
 
OT 5101 week2 theory policy
OT 5101 week2 theory policyOT 5101 week2 theory policy
OT 5101 week2 theory policy
 
OT 5101 week3 planning needs assessment
OT 5101 week3 planning needs assessmentOT 5101 week3 planning needs assessment
OT 5101 week3 planning needs assessment
 
Ot5101 week1
Ot5101 week1Ot5101 week1
Ot5101 week1
 
NUR 304 Chapter005
NUR 304 Chapter005NUR 304 Chapter005
NUR 304 Chapter005
 
NUR 3043 Chapter007
NUR 3043 Chapter007NUR 3043 Chapter007
NUR 3043 Chapter007
 
NUR 3043 Chapter006
NUR 3043 Chapter006NUR 3043 Chapter006
NUR 3043 Chapter006
 
NUR 3043 Chapter004
NUR 3043 Chapter004NUR 3043 Chapter004
NUR 3043 Chapter004
 
3043 Chapter009
3043 Chapter0093043 Chapter009
3043 Chapter009
 
3043 Chapter008
 3043 Chapter008 3043 Chapter008
3043 Chapter008
 
Melnyk ppt chapter_21
Melnyk ppt chapter_21Melnyk ppt chapter_21
Melnyk ppt chapter_21
 
Melnyk ppt chapter_22
Melnyk ppt chapter_22Melnyk ppt chapter_22
Melnyk ppt chapter_22
 

Recently uploaded

Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...EduSkills OECD
 
Alper Gobel In Media Res Media Component
Alper Gobel In Media Res Media ComponentAlper Gobel In Media Res Media Component
Alper Gobel In Media Res Media ComponentInMediaRes1
 
A Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformA Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformChameera Dedduwage
 
The basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptxThe basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptxheathfieldcps1
 
SOCIAL AND HISTORICAL CONTEXT - LFTVD.pptx
SOCIAL AND HISTORICAL CONTEXT - LFTVD.pptxSOCIAL AND HISTORICAL CONTEXT - LFTVD.pptx
SOCIAL AND HISTORICAL CONTEXT - LFTVD.pptxiammrhaywood
 
MENTAL STATUS EXAMINATION format.docx
MENTAL     STATUS EXAMINATION format.docxMENTAL     STATUS EXAMINATION format.docx
MENTAL STATUS EXAMINATION format.docxPoojaSen20
 
Accessible design: Minimum effort, maximum impact
Accessible design: Minimum effort, maximum impactAccessible design: Minimum effort, maximum impact
Accessible design: Minimum effort, maximum impactdawncurless
 
mini mental status format.docx
mini    mental       status     format.docxmini    mental       status     format.docx
mini mental status format.docxPoojaSen20
 
Sanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdfSanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdfsanyamsingh5019
 
Separation of Lanthanides/ Lanthanides and Actinides
Separation of Lanthanides/ Lanthanides and ActinidesSeparation of Lanthanides/ Lanthanides and Actinides
Separation of Lanthanides/ Lanthanides and ActinidesFatimaKhan178732
 
The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13Steve Thomason
 
Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)eniolaolutunde
 
Incoming and Outgoing Shipments in 1 STEP Using Odoo 17
Incoming and Outgoing Shipments in 1 STEP Using Odoo 17Incoming and Outgoing Shipments in 1 STEP Using Odoo 17
Incoming and Outgoing Shipments in 1 STEP Using Odoo 17Celine George
 
“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...
“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...
“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...Marc Dusseiller Dusjagr
 
Solving Puzzles Benefits Everyone (English).pptx
Solving Puzzles Benefits Everyone (English).pptxSolving Puzzles Benefits Everyone (English).pptx
Solving Puzzles Benefits Everyone (English).pptxOH TEIK BIN
 
microwave assisted reaction. General introduction
microwave assisted reaction. General introductionmicrowave assisted reaction. General introduction
microwave assisted reaction. General introductionMaksud Ahmed
 
Enzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdf
Enzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdfEnzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdf
Enzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdfSumit Tiwari
 
How to Make a Pirate ship Primary Education.pptx
How to Make a Pirate ship Primary Education.pptxHow to Make a Pirate ship Primary Education.pptx
How to Make a Pirate ship Primary Education.pptxmanuelaromero2013
 
KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...
KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...
KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...M56BOOKSTORE PRODUCT/SERVICE
 

Recently uploaded (20)

Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
Presentation by Andreas Schleicher Tackling the School Absenteeism Crisis 30 ...
 
Alper Gobel In Media Res Media Component
Alper Gobel In Media Res Media ComponentAlper Gobel In Media Res Media Component
Alper Gobel In Media Res Media Component
 
A Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformA Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy Reform
 
The basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptxThe basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptx
 
SOCIAL AND HISTORICAL CONTEXT - LFTVD.pptx
SOCIAL AND HISTORICAL CONTEXT - LFTVD.pptxSOCIAL AND HISTORICAL CONTEXT - LFTVD.pptx
SOCIAL AND HISTORICAL CONTEXT - LFTVD.pptx
 
MENTAL STATUS EXAMINATION format.docx
MENTAL     STATUS EXAMINATION format.docxMENTAL     STATUS EXAMINATION format.docx
MENTAL STATUS EXAMINATION format.docx
 
Accessible design: Minimum effort, maximum impact
Accessible design: Minimum effort, maximum impactAccessible design: Minimum effort, maximum impact
Accessible design: Minimum effort, maximum impact
 
mini mental status format.docx
mini    mental       status     format.docxmini    mental       status     format.docx
mini mental status format.docx
 
Sanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdfSanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdf
 
Separation of Lanthanides/ Lanthanides and Actinides
Separation of Lanthanides/ Lanthanides and ActinidesSeparation of Lanthanides/ Lanthanides and Actinides
Separation of Lanthanides/ Lanthanides and Actinides
 
The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13
 
Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)Software Engineering Methodologies (overview)
Software Engineering Methodologies (overview)
 
Incoming and Outgoing Shipments in 1 STEP Using Odoo 17
Incoming and Outgoing Shipments in 1 STEP Using Odoo 17Incoming and Outgoing Shipments in 1 STEP Using Odoo 17
Incoming and Outgoing Shipments in 1 STEP Using Odoo 17
 
Model Call Girl in Bikash Puri Delhi reach out to us at 🔝9953056974🔝
Model Call Girl in Bikash Puri  Delhi reach out to us at 🔝9953056974🔝Model Call Girl in Bikash Puri  Delhi reach out to us at 🔝9953056974🔝
Model Call Girl in Bikash Puri Delhi reach out to us at 🔝9953056974🔝
 
“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...
“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...
“Oh GOSH! Reflecting on Hackteria's Collaborative Practices in a Global Do-It...
 
Solving Puzzles Benefits Everyone (English).pptx
Solving Puzzles Benefits Everyone (English).pptxSolving Puzzles Benefits Everyone (English).pptx
Solving Puzzles Benefits Everyone (English).pptx
 
microwave assisted reaction. General introduction
microwave assisted reaction. General introductionmicrowave assisted reaction. General introduction
microwave assisted reaction. General introduction
 
Enzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdf
Enzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdfEnzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdf
Enzyme, Pharmaceutical Aids, Miscellaneous Last Part of Chapter no 5th.pdf
 
How to Make a Pirate ship Primary Education.pptx
How to Make a Pirate ship Primary Education.pptxHow to Make a Pirate ship Primary Education.pptx
How to Make a Pirate ship Primary Education.pptx
 
KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...
KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...
KSHARA STURA .pptx---KSHARA KARMA THERAPY (CAUSTIC THERAPY)————IMP.OF KSHARA ...
 

Math3010 week 5

  • 1. ReviewReview  Types of dataTypes of data  p-valuep-value  Steps for hypothesis testSteps for hypothesis test – How do we set up a null hypothesis?How do we set up a null hypothesis?  Choosing the right testChoosing the right test – Continuous outcome variable/dichotomousContinuous outcome variable/dichotomous explanatory variable: Two sample t-testexplanatory variable: Two sample t-test
  • 2. Steps for hypothesis testingSteps for hypothesis testing 1)1) State null hypothesisState null hypothesis 2)2) State type of data for explanatory and outcomeState type of data for explanatory and outcome variablevariable 3)3) Determine appropriate statistical testDetermine appropriate statistical test 4)4) State summary statisticsState summary statistics 5)5) Calculate p-value (stat package)Calculate p-value (stat package) 6)6) Decide whether to reject or not reject the nullDecide whether to reject or not reject the null hypothesishypothesis • NEVER accept nullNEVER accept null 1)1) Write conclusionWrite conclusion
  • 3. ExampleExample  In previous class, two groups wereIn previous class, two groups were compared on a continuous outcomecompared on a continuous outcome  What if we have more than two groups?What if we have more than two groups?  Ex. A recent study compared the intensityEx. A recent study compared the intensity of structures on MRI in normal controls,of structures on MRI in normal controls, benign MS patients and secondarybenign MS patients and secondary progressive MS patientsprogressive MS patients  Question: Is there any difference amongQuestion: Is there any difference among these groups?these groups?
  • 4.
  • 5. Two approachesTwo approaches  Compare each group to each other groupCompare each group to each other group using a t-testusing a t-test – Problem withProblem with multiple comparisonsmultiple comparisons  CompleteComplete global comparisonglobal comparison to see ifto see if there is any differencethere is any difference – Analysis of variance (ANOVA)Analysis of variance (ANOVA) – Good first step even if eventually completeGood first step even if eventually complete pairwise comparisonspairwise comparisons
  • 6. Types of analysis-independentTypes of analysis-independent samplessamples OutcomeOutcome ExplanatoryExplanatory AnalysisAnalysis ContinuousContinuous DichotomousDichotomous t-test, Wilcoxont-test, Wilcoxon testtest ContinuousContinuous CategoricalCategorical ANOVA, linearANOVA, linear regressionregression ContinuousContinuous ContinuousContinuous Correlation, linearCorrelation, linear regressionregression DichotomousDichotomous DichotomousDichotomous Chi-square test,Chi-square test, logistic regressionlogistic regression DichotomousDichotomous ContinuousContinuous Logistic regressionLogistic regression Time to eventTime to event DichotomousDichotomous Log-rank testLog-rank test
  • 7. Global test-ANOVAGlobal test-ANOVA  As a first step, we can compare across allAs a first step, we can compare across all groups at oncegroups at once  The null hypothesis for ANOVA is that theThe null hypothesis for ANOVA is that the means in all of the groups are equalmeans in all of the groups are equal  ANOVA compares the within groupANOVA compares the within group variance and the between group variancevariance and the between group variance – If the patients within a group are very alikeIf the patients within a group are very alike and the groups are very different, the groupsand the groups are very different, the groups are likely differentare likely different
  • 8.
  • 9. Hypothesis testHypothesis test 1)1) HH00: mean: meannormalnormal=mean=meanBMSBMS=mean=meanSPMSSPMS 2)2) Outcome variable: continuousOutcome variable: continuous Explanatory variable: categoricalExplanatory variable: categorical 3)3) Test: ANOVATest: ANOVA 4)4) meanmeannormalnormal=0.41; mean=0.41; meanBMSBMS= 0.34; mean= 0.34; meanSPMSSPMS=0.30=0.30 5)5) Results: p=0.011Results: p=0.011 6)6) Reject null hypothesisReject null hypothesis 7)7) Conclusion: At least one of the groups isConclusion: At least one of the groups is significantly different than the otherssignificantly different than the others
  • 10.
  • 11. Technical asideTechnical aside  Our F-statistic is the ratio of the between groupOur F-statistic is the ratio of the between group variance and the within group variancevariance and the within group variance  This ratio of variances has a known distribution (F-This ratio of variances has a known distribution (F- distribution)distribution)  If our calculated F-statistic is high, the between groupIf our calculated F-statistic is high, the between group variance is higher than the within group variance,variance is higher than the within group variance, meaning the differences between the groups are notmeaning the differences between the groups are not likely due to chancelikely due to chance  Therefore, the probability of the observed result orTherefore, the probability of the observed result or something more extreme will be low (low p-value)something more extreme will be low (low p-value) ( ) ( ) ( )( ) ( ) ( )( )1111 1 1 22 11 1 2 2 2 −++−−++− −− == ∑= kkk k i ii within between nnsnsn kxxn s s F 
  • 12. This is the distribution under the null This small shaded region is the part of the distribution that is equal to or more extreme than the observed value. The p-value!!!
  • 13. Now whatNow what  The question often becomes which groupsThe question often becomes which groups are differentare different  Possible comparisonsPossible comparisons – All pairsAll pairs – All groups to a specific controlAll groups to a specific control – Pre-specified comparisonsPre-specified comparisons  If we do many tests, we should account forIf we do many tests, we should account for multiple comparisonsmultiple comparisons
  • 14. Type I errorType I error  Type I error is when you reject the nullType I error is when you reject the null hypothesis even though it is truehypothesis even though it is true ((αα=P(reject H=P(reject H00|H|H00 is true))is true))  We accept making this error 5% of theWe accept making this error 5% of the timetime  If we run a large experiment with 100 testsIf we run a large experiment with 100 tests and the null hypothesis was true in eachand the null hypothesis was true in each case, how many times would we expect tocase, how many times would we expect to reject the null?reject the null?
  • 15. Multiple comparisonsMultiple comparisons  For this problem, three comparisonsFor this problem, three comparisons – NC vs. BMS; NC vs. SPMS; BMS vs. SPMSNC vs. BMS; NC vs. SPMS; BMS vs. SPMS  If we complete each test at the 0.05 level, whatIf we complete each test at the 0.05 level, what is the chance that we make a type I error?is the chance that we make a type I error? – P(reject at least 1 | HP(reject at least 1 | H00 is true)is true) == αα – P(reject at least 1 | HP(reject at least 1 | H00 is true)is true) = 1-= 1- P(fail to reject allP(fail to reject all three| Hthree| H00 is true)is true) = 1-0.95= 1-0.9533 = 0.143= 0.143  Inflated type I error rateInflated type I error rate  Can correct p-value for each test to maintainCan correct p-value for each test to maintain experiment type I errorexperiment type I error
  • 16. Bonferroni correctionBonferroni correction  TheThe Bonferroni correctionBonferroni correction multiples all p-multiples all p- values by the number of comparisons completedvalues by the number of comparisons completed – In our experiment, there were 3 comparisons, so weIn our experiment, there were 3 comparisons, so we multiply by 3multiply by 3 – Any p-value that remains less than 0.05 is significantAny p-value that remains less than 0.05 is significant  The Bonferroni correction is conservative (it isThe Bonferroni correction is conservative (it is more difficult to obtain a significant result than itmore difficult to obtain a significant result than it should be), but it is an extremely easy way toshould be), but it is an extremely easy way to account for multiple comparisons.account for multiple comparisons. – Can be very harsh correction with many testsCan be very harsh correction with many tests
  • 17. Other correctionsOther corrections  All pairwise comparisonsAll pairwise comparisons – Tukey’s testTukey’s test  All groups to a controlAll groups to a control – Dunnett’s testDunnett’s test  MANY othersMANY others  False discovery rateFalse discovery rate
  • 18. ExampleExample  For our three-group comparison, we compareFor our three-group comparison, we compare each and get the following results from Tukey’seach and get the following results from Tukey’s testtest GroupsGroups Mean diffMean diff p-valuep-value SignificantSignificant NC vs. BMSNC vs. BMS 0.0750.075 0.100.10 NC vs. SPMSNC vs. SPMS 0.1140.114 0.0120.012 ** BMS vs.BMS vs. SPMSSPMS 0.0390.039 0.600.60
  • 19.
  • 20. Questions to ask yourselfQuestions to ask yourself  What is the null hypothesis?What is the null hypothesis?  We would like to test the null hypothesis atWe would like to test the null hypothesis at the 0.05 levelthe 0.05 level  If well defined prior to the experiment, theIf well defined prior to the experiment, the correction for multiple comparison ifcorrection for multiple comparison if necessary will be clearnecessary will be clear  Hypothesis generating vs.Hypothesis generating vs. hypothesis testinghypothesis testing
  • 21. ConclusionsConclusions  If you are doing a multiple group comparison,If you are doing a multiple group comparison, always specify before the experiment whichalways specify before the experiment which comparisons are of interest if possiblecomparisons are of interest if possible  If the null hypothesis is that all the groups areIf the null hypothesis is that all the groups are the same, test global null using ANOVAthe same, test global null using ANOVA  Complete appropriate additional comparisonsComplete appropriate additional comparisons with corrections if necessarywith corrections if necessary  No single right answer for every situationNo single right answer for every situation
  • 22. Types of analysis-independentTypes of analysis-independent samplessamples OutcomeOutcome ExplanatoryExplanatory AnalysisAnalysis ContinuousContinuous DichotomousDichotomous t-test, Wilcoxont-test, Wilcoxon testtest ContinuousContinuous CategoricalCategorical ANOVA, linearANOVA, linear regressionregression ContinuousContinuous ContinuousContinuous Correlation, linearCorrelation, linear regressionregression DichotomousDichotomous DichotomousDichotomous Chi-square test,Chi-square test, logistic regressionlogistic regression DichotomousDichotomous ContinuousContinuous Logistic regressionLogistic regression Time to eventTime to event DichotomousDichotomous Log-rank testLog-rank test
  • 23. CorrelationCorrelation  Is there a linearIs there a linear relationship betweenrelationship between IL-10 expression andIL-10 expression and IL-6 expression?IL-6 expression?  The best graphicalThe best graphical display for this data isdisplay for this data is a scatter plota scatter plot
  • 24. CorrelationCorrelation  DefinitionDefinition: the degree to which two continuous: the degree to which two continuous variables are linearly relatedvariables are linearly related – Positive correlation- As one variable goes up, thePositive correlation- As one variable goes up, the other goes up (positive slope)other goes up (positive slope) – Negative correlation- As one variable goes up, theNegative correlation- As one variable goes up, the other goes down (negative slope)other goes down (negative slope)  Correlation (Correlation (ρρ) ranges from -1 (perfect negative) ranges from -1 (perfect negative correlation) to 1 (perfect positive correlation)correlation) to 1 (perfect positive correlation)  A correlation of 0 means that there is no linearA correlation of 0 means that there is no linear relationship between the two variablesrelationship between the two variables
  • 25. Positive correlation 0 2 4 6 8 10 12 0 2 4 6 8 10 12 Negative correlation 0 2 4 6 8 10 12 0 2 4 6 8 10 12 No correlation 0 1 2 3 4 5 6 7 8 9 10 0 2 4 6 8 10 12 No correlation (quadratic) 0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10
  • 26. Hypothesis testHypothesis test 1)1) HH00: correlation between IL-10 expression and: correlation between IL-10 expression and IL-6 expression=0IL-6 expression=0 2)2) Outcome variable: IL-6 expression- continuousOutcome variable: IL-6 expression- continuous Explanatory variable: IL-10 expression-Explanatory variable: IL-10 expression- continuouscontinuous 3)3) Test: correlationTest: correlation 4)4) Summary statistic: correlation=0.51Summary statistic: correlation=0.51 5)5) Results: p=0.011Results: p=0.011 6)6) Reject null hypothesisReject null hypothesis 7)7) Conclusion: A statistically significantConclusion: A statistically significant correlation was observed between the twocorrelation was observed between the two variablesvariables
  • 27.
  • 28. Technical aside-correlationTechnical aside-correlation  The formal definition of the correlation is given by:The formal definition of the correlation is given by:  Note that this is dimensionless quantityNote that this is dimensionless quantity  This equation shows that if the covariance between theThis equation shows that if the covariance between the two variables is the same as the variance in the twotwo variables is the same as the variance in the two variables, we have perfect correlation because all of thevariables, we have perfect correlation because all of the variability in x and y is explained by how the twovariability in x and y is explained by how the two variables change togethervariables change together )()( ),( ),( yVarxVar yxCov yxCorr =
  • 29. How can we estimate theHow can we estimate the correlation?correlation?  The most common estimator of the correlation is theThe most common estimator of the correlation is the Pearson’s correlation coefficientPearson’s correlation coefficient, given by:, given by:  This is a estimate that requires both x and y are normallyThis is a estimate that requires both x and y are normally distributed. Since we use the mean in the calculation, thedistributed. Since we use the mean in the calculation, the estimate is sensitive to outliers.estimate is sensitive to outliers. ( )( ) ( ) ( )       −      − −− = ∑∑ ∑ == = n i i n i i n i ii yyxx yyxx r 1 2 1 2 1
  • 30. Distribution of the test statisticDistribution of the test statistic  The standard error of the sample correlationThe standard error of the sample correlation coefficient is given bycoefficient is given by  The resulting distribution of the test statistic is a t-The resulting distribution of the test statistic is a t- distribution with n-2 degrees of freedom where ndistribution with n-2 degrees of freedom where n is the number of patients (not the number ofis the number of patients (not the number of measurements)measurements) 2 1 )(ˆ 2 − − = n r res 22 1 2 2 1 0 r n r n r r t − − = − − − =
  • 31. Regression-Everything in one placeRegression-Everything in one place  All analyses we have done to this pointAll analyses we have done to this point can be completed using regression!!!can be completed using regression!!!
  • 32. Quick math reviewQuick math review  As you remember, theAs you remember, the equation of a line isequation of a line is y=mx+by=mx+b  FFor every one unitor every one unit increase in x, there isincrease in x, there is an m unit increase inan m unit increase in yy  bb is the value of yis the value of y when x is equal towhen x is equal to zerozero Line y = 1.5x + 4 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12
  • 33. PicturePicture  Does there seem toDoes there seem to be a linearbe a linear relationship in therelationship in the data?data?  Is the data perfectlyIs the data perfectly linear?linear?  Could we fit a line toCould we fit a line to this data?this data? 0 5 10 15 20 25 0 2 4 6 8 10 12
  • 34. How do we find the best line?How do we find the best line?  Linear regressionLinear regression tries to find the besttries to find the best line (curve) to fit theline (curve) to fit the data Let’s look atdata Let’s look at three candidate linesthree candidate lines  Which do you think isWhich do you think is the best?the best?  What is a way toWhat is a way to determine the bestdetermine the best line to use?line to use?
  • 35. What is linear regression?What is linear regression?  The method of findingThe method of finding the best line (curve) isthe best line (curve) is least squares, whichleast squares, which minimizes theminimizes the distance from the linedistance from the line for each of pointsfor each of points  The equation of theThe equation of the line is y=1.5x + 4line is y=1.5x + 4 y = 1.5x + 4 0 5 10 15 20 25 0 2 4 6 8 10 12
  • 36. ExampleExample  For our investigation of theFor our investigation of the relationship between IL-10relationship between IL-10 and IL-6, we can set up aand IL-6, we can set up a regression equationregression equation  ββ00 is the expression of IL-6is the expression of IL-6 when IL-10=0 (intercept)when IL-10=0 (intercept)  ββ11 is the change in IL-6 foris the change in IL-6 for every 1 unit increase in IL-10every 1 unit increase in IL-10 (slope)(slope)  εεii is the residual from the lineis the residual from the line iii ILIL εββ ++= 10*6 10
  • 37.  The final regression equation isThe final regression equation is  The coefficients meanThe coefficients mean – the estimate of the mean expression of IL-6 for athe estimate of the mean expression of IL-6 for a patient with IL-10 expression=0 (patient with IL-10 expression=0 (ββ00)) – an increase of one unit in IL-10 expression leads toan increase of one unit in IL-10 expression leads to an estimated increase of 0.63 in the meanan estimated increase of 0.63 in the mean expression of IL-6 (expression of IL-6 (ββ11)) 10*63.04.266ˆ ILIL +=
  • 38. Tough questionTough question  In our correlation hypothesis test, we wanted toIn our correlation hypothesis test, we wanted to know if there was an association between theknow if there was an association between the two measurestwo measures  If there was no relationship between IL-10 andIf there was no relationship between IL-10 and IL-6 in our system, what would happen to ourIL-6 in our system, what would happen to our regression equation?regression equation? – No effect means that the change in IL-6 is not relatedNo effect means that the change in IL-6 is not related to the change in IL-10to the change in IL-10 – ββ11=0=0  IsIs ββ11 significantly different than zero?significantly different than zero?
  • 39. Hypothesis testHypothesis test 1)1) HH00: no relationship between IL-6 expression: no relationship between IL-6 expression and IL-10 expression,and IL-10 expression, ββ11 =0=0 2)2) Outcome variable: IL-6- continuousOutcome variable: IL-6- continuous Explanatory variable: IL-10- continuousExplanatory variable: IL-10- continuous 3)3) Test: linear regressionTest: linear regression 4)4) Summary statistic:Summary statistic: ββ11 = 0.63= 0.63 5)5) Results: p=0.011Results: p=0.011 6)6) Reject null hypothesisReject null hypothesis 7)7) Conclusion: A significant correlation wasConclusion: A significant correlation was observed between the two variablesobserved between the two variables
  • 40.
  • 41. Wait a second!!Wait a second!!  Let’s check somethingLet’s check something – p-value from correlation analysis = 0.011p-value from correlation analysis = 0.011 – p-value from regression analysis = 0.011p-value from regression analysis = 0.011 – They are the same!!They are the same!!  Regression leads to same conclusion asRegression leads to same conclusion as correlation analysiscorrelation analysis  Other similarities as well from modelsOther similarities as well from models
  • 42. Technical aside-Estimates ofTechnical aside-Estimates of regression coefficientsregression coefficients  Once we have solved the least squaresOnce we have solved the least squares equation, we obtain estimates for theequation, we obtain estimates for the ββ’s, which’s, which we refer to aswe refer to as  To test if this estimate is significantly differentTo test if this estimate is significantly different than 0, we use the following equation:than 0, we use the following equation: 10 ˆ,ˆ ββ ( )( ) ( ) xy xx yyxx n i i n i ii 10 1 2 1 1 ˆˆ ˆ ββ β −= − −− = ∑ ∑ = = ( )1 11 ˆˆ ˆ β ββ es t − =
  • 43. Assumptions of linear regressionAssumptions of linear regression  LinearityLinearity – Linear relationship between outcome and predictorsLinear relationship between outcome and predictors – E(Y|X=x)=E(Y|X=x)=ββ00 ++ ββ11xx11 ++ ββ22xx22 22 is still a linear regressionis still a linear regression equation because each of theequation because each of the ββ’s is to the first power’s is to the first power  Normality of the residualsNormality of the residuals – The residuals,The residuals, εεii, are normally distributed, N(0,, are normally distributed, N(0, σσ22 ))  Homoscedasticity of the residualsHomoscedasticity of the residuals – The residuals,The residuals, εεii, have the same variance, have the same variance  IndependenceIndependence – All of the data points are independentAll of the data points are independent – Correlated data points can be taken into accountCorrelated data points can be taken into account using multivariate and longitudinal data methodsusing multivariate and longitudinal data methods
  • 44. Linear regression with dichotomousLinear regression with dichotomous predictorpredictor  Linear regression can also be used forLinear regression can also be used for dichotomous predictors, like sexdichotomous predictors, like sex  Last class we compared relapsing MS patientsLast class we compared relapsing MS patients to progressive MS patientsto progressive MS patients  To do this, we use an indicator variable, whichTo do this, we use an indicator variable, which equals 1 for relapsing and 0 for progressive. Theequals 1 for relapsing and 0 for progressive. The resulting regression equation for expression isresulting regression equation for expression is iii Rex εββ ++= *10
  • 45. Interpretation of modelInterpretation of model  The meaning of the coefficients in this case areThe meaning of the coefficients in this case are – ββ00 is the estimate of the mean expression whenis the estimate of the mean expression when R=0, in the progressive groupR=0, in the progressive group – ββ00 + β+ β11 is the estimate of the mean expression whenis the estimate of the mean expression when R=1, in the relapsing groupR=1, in the relapsing group – ββ11 is the estimate of the mean increase in expressionis the estimate of the mean increase in expression between the two groupsbetween the two groups  The difference between the two groups isThe difference between the two groups is ββ11  If there was no difference between the groups,If there was no difference between the groups, what wouldwhat would ββ11 equal?equal?
  • 46. Mean in wildtype=β0 Mean in Progressive group=β0 Difference between groups=β1
  • 47. Hypothesis testHypothesis test 1)1) Null hypothesis: meanNull hypothesis: meanprogressiveprogressive=mean=meanrelapsingrelapsing ((ββ11=0)=0) 2)2) Explanatory: group membership- dichotomousExplanatory: group membership- dichotomous Outcome: cytokine production-continuousOutcome: cytokine production-continuous 3)3) Test: Linear regressionTest: Linear regression 4)4) ββ11=6.87=6.87 5)5) p-value=0.199p-value=0.199 6)6) Fail to reject null hypothesisFail to reject null hypothesis 7)7) Conclusion: The difference between theConclusion: The difference between the groups is not statistically significantgroups is not statistically significant
  • 48. T-testT-test  As hopefully you remember, you couldAs hopefully you remember, you could have tested this same null hypothesishave tested this same null hypothesis using a two sample t-testusing a two sample t-test  Very similar result to previous classVery similar result to previous class  If we would have assumed equal varianceIf we would have assumed equal variance for our t-test, we would have gotten to thefor our t-test, we would have gotten to the same result!!!same result!!!  ANOVA results can also be tested usingANOVA results can also be tested using regression using more than one indicatorregression using more than one indicator
  • 49. Multiple regressionMultiple regression  A large advantage of regression is the ability toA large advantage of regression is the ability to include multiple predictors of an outcome in oneinclude multiple predictors of an outcome in one analysisanalysis  A multiple regression equation looks just like aA multiple regression equation looks just like a simple regression equation.simple regression equation. exxxY nn +++++= ββββ ...22110
  • 50. ExampleExample  Brain parenchymal fraction (BPF) is aBrain parenchymal fraction (BPF) is a measure of disease severity in MSmeasure of disease severity in MS  We would like to know if gender has anWe would like to know if gender has an effect on BPF in MS patientseffect on BPF in MS patients  We also know that BPF declines with ageWe also know that BPF declines with age in MS patientsin MS patients  Is there an effect of sex on BPF if weIs there an effect of sex on BPF if we control for age?control for age?
  • 51. .75.8.85.9.95 BPF 0 .2 .4 .6 .8 1 Sex Blue=males; Red=females
  • 53. Is age a potential confounder?Is age a potential confounder?  We know that age has an effect on BPFWe know that age has an effect on BPF from previous researchfrom previous research  We also know that male patients have aWe also know that male patients have a different disease course than femaledifferent disease course than female patients so the age at time of samplingpatients so the age at time of sampling may also be related to sexmay also be related to sex BPFSex Age
  • 54. ModelModel  The multiple linear regression modelThe multiple linear regression model includes a term for both age and sexincludes a term for both age and sex  What are the values genderWhat are the values genderii takes on?takes on? – gendergenderii=0 if the patient is female=0 if the patient is female – gendergenderii=1 if the patient is male=1 if the patient is male iiii agegenderBPF εβββ +++= ** 210
  • 55. ExpressionExpression  Females:Females: – BPFBPFii == ββ00++ ββ22*age*ageii++εεii  Males:Males: – BPFBPFii = (= (ββ00++ ββ11)+)+ ββ22*age*ageii++εεii  What is different about the equations?What is different about the equations? – InterceptIntercept  What is the same?What is the same? – SlopeSlope  This model allows an effect of gender on theThis model allows an effect of gender on the intercept, but not on the change with ageintercept, but not on the change with age
  • 56.  The meaning of each coefficientThe meaning of each coefficient – ββ00:: the average BPF when age is 0 and the patient isthe average BPF when age is 0 and the patient is femalefemale – ββ11:: the average difference in BPF between males andthe average difference in BPF between males and female, HOLDING AGE CONSTANTfemale, HOLDING AGE CONSTANT – ββ22:: the average increase in BPF for a one unitthe average increase in BPF for a one unit increase in age, HOLDING GENDER CONSTANTincrease in age, HOLDING GENDER CONSTANT  Note that the interpretation of the coefficientNote that the interpretation of the coefficient requires mention of the other variables in therequires mention of the other variables in the modelmodel Interpretation of coefficientsInterpretation of coefficients
  • 57. Estimated coefficientsEstimated coefficients  Here is the estimated regression equationHere is the estimated regression equation  The average difference between males andThe average difference between males and females is 0.017 holding age constantfemales is 0.017 holding age constant  For every one unit increase in age, the meanFor every one unit increase in age, the mean BPF decreases 0.0026 units holding sex constantBPF decreases 0.0026 units holding sex constant  Are either of these effects statistically significant?Are either of these effects statistically significant? – What is the null hypothesis?What is the null hypothesis? iii agesexFBP *0026.0*017.0942.0ˆ −+=
  • 58. Hypothesis testHypothesis test 1)1) HH00: No effect of sex, controlling for age: No effect of sex, controlling for age ββ11 =0=0 2)2) Continuous outcome, continuous predictorContinuous outcome, continuous predictor 3)3) Linear regression controlling for sexLinear regression controlling for sex 4)4) Summary statistic:Summary statistic: ββ11 =0.017=0.017 5)5) p-value=0.37p-value=0.37 6)6) Since the p-value is more than 0.05, we fail toSince the p-value is more than 0.05, we fail to reject the null hypothesisreject the null hypothesis 7)7) We conclude that there is no significantWe conclude that there is no significant association between sex and BPF controllingassociation between sex and BPF controlling for agefor age
  • 59. Hypothesis testHypothesis test 1)1) HH00: No effect of age, controlling for sex: No effect of age, controlling for sex ββ22 =0=0 2)2) Continuous outcome, continuous predictorContinuous outcome, continuous predictor 3)3) Linear regression controlling for sexLinear regression controlling for sex 4)4) Summary statistic:Summary statistic: ββ22 =-0.0026=-0.0026 5)5) p-value=0.00p-value=0.00 44 6)6) Since the p-value is less than 0.05, we rejectSince the p-value is less than 0.05, we reject the null hypothesisthe null hypothesis 7)7) We conclude that there is a significantWe conclude that there is a significant association between age and BPF controllingassociation between age and BPF controlling for sexfor sex
  • 60. Estimated effect of age p-value for age Estimated effect of sex p-value for sex
  • 62. ConclusionsConclusions  Although there was a marginallyAlthough there was a marginally significant association of sex and BPF,significant association of sex and BPF, this association was not significant afterthis association was not significant after controlling for agecontrolling for age  The significant association between ageThe significant association between age and BPF remained statistically significantand BPF remained statistically significant after controlling for sexafter controlling for sex
  • 63. What we learned (hopefully)What we learned (hopefully)  ANOVAANOVA  CorrelationCorrelation  Basics of regressionBasics of regression